Fig. 1.

Overall study design. (A) Trajectory of a COVID-19 patient transitioning from pre-COVID-19 (time up to 365 days before first positive PCR test) through the SARS-CoV-2 positive phase (interval after first positive test but before the first of two consecutive negative PCR test results) into the post viral clearance phase (period up to 90 days after the first of two negative PCR test results), (B) Demonstrates procedure for assigning patients to the hospitalized post-clearance cohort and non-hospitalized post-clearance cohort patients can be hospitalized at varying points in time including during the index infection (time from first positive PCR test results to first of two negative PCR test results) and following viral clearance - two cohorts are defined from the overall population, Hospitalized Post-Clearance Cohort in which patients are admitted or readmitted to the hospital following their estimated clearance date and Non-Hospitalized Post-Clearance Cohort in which patients are admitted during the index infection, but not following the estimated clearance date, (C) For each patient if the two defined cohorts a deep language (BERT) model is used to extracted phenotypes of interest from the clinical notes recorded between 365 days prior to infection and up to 90 days after clearance for each patient - occurrences of these phenotypes are stratified into pre-COVID-19, COVID-19 pre-clearance, and COVID-19 post-clearance time periods and statistical tests are run to find significant differences in phenotypes between the two cohorts.